Image Representation of Acoustic Features for the Automatic Recognition of Underwater Noise Targets

Feature extraction is one of the most important technologies for underwater targets recognition. In the past few decades, a number of methods for feature extraction have been developed, and under certain conditions they can achieve high recognition rate. However, for complex environments, it is still difficult to improve the robustness of the recognition system, and new robust feature extraction methods are expectant. This paper presents a novel method of feature extraction based on the spectrogram of acoustic signals. The image moment features and image texture features are extracted and the algorithms of LDA, PCA and their combinations are used to select the effective features respectively. The experimental results show that, these selected image features can achieve high recognition rate.